The long term goal of this research is to develop an effective and extensible computational model ofHIV drug therapy to help in fighting the increasingly prevalent problem of drug resistance in. Utilizing ahierarchy of methods that range from atomic level molecular co-evolution to bioinformatics analysis of patientdata, we will characterize the mutational landscape available to the virus under drug selection pressure anddevelop structure-based design methods that can produce inhibitors and therapeutic strategies to efficientlydefeat escape. Our work will focus primarily on the well-characterized HIV protease, but will extend to otherHIV drug targets to test the generality of our methods. In this research we will evaluate two basichypotheses:1) There exists a relatively small number of definable classes of HIV protease mutations, such thatmutants within a given class will show strong cross resistance to a given inhibitor, and mutants indifferent classes wil not show cross-resistance.2) Fragment-based design combined with computational modeling of protein flexibility will identifyinhibitors that expand the protease 'target space' including identification of exosites and design offlexibility wedges that block critical functional motions of HIV protease.Working in conjunction with the fragment-based crystallographic studies undertaken in Project 2 and thecombinatorial chemical syntheses in Project 3 we will develop and apply computational methods forfragment-based drug design. Utilizing dynamic models of HIV protease wild type and mutants ourFightAIDS@Home Internet distributed computing network will provide the needed computational power toscreen large fragment libraries and evaluate dockings of promising linked fragments. Based on combinationof experimental results on ex vivo resistance evolution from Project 4 and time course patient treatment datafrom Core C and by running and analyzing massive computational coevolution experiments, we willcharacterize the space of possible mutants, separating the range of possible viable mutants into discretestructural classes and identifying key mutant structures within each class for use in drug design. We willdevelop models of viral fitness that incorporate these multiple sources of data, and then use these modelswithin larger simulations of viral evolution during the course of drug therapy.
Showing the most recent 10 out of 28 publications
